AI-Cancer-Detection Project Using CNN

Authors : Amanda Loeung, Hamza Harb, Melodie Cornelly

Instructors : Dr. Christelle Scharff , Dr. Kaleema, Stephanie Sicilian

Data set used from Kaggle - https://www.kaggle.com/datasets/indk214/brain-tumor-dataset-segmentation-and-classification

Project Objective

This project aims to assist medical professionals in detecting brain tumors from MRI scans using deep learning. It explores the effectiveness of CNNs in medical image classification and serves as a reference for future diagnostic tools.

Project Structure

Data Download: From Kaggle using kagglehub

• Data Preprocessing: Rescaling, rotation, zoom, and brightness adjustment
• Model Architecture: CNN (or EfficientNet if applied), trained on 224x224 MRI images
• Training: Augmented data, validation split, early stopping, and checkpointing
• Evaluation: Accuracy score, confusion matrix, classification report
• Prediction: Model can predict on new/unseen brain MRI scans

Tools Used

• Python 3
• TensorFlow / Keras
• NumPy, Matplotlib
• KaggleHub
• Jupyter Notebook

Data Augmentation Techniques

To improve generalization we applied the following :

• rotation_range=10 – minor rotations
• width/height_shift=0.05 – shift tumor regions
• zoom_range=0.15 – vary tumor sizes
• brightness_range=[0.8, 1.2] – simulate scanning variations
• horizontal_flip=False – preserved anatomical orientation
• validation_split=0.2 - helps monitor how well the model generalizes to unseen data

Model Performance

• Total Images: 7,023 (5,712 training / 1,311 testing)
• Tumor Classes: 4 (e.g., Glioma, Meningioma, Pituitary, No Tumor)
• Epochs Trained: 5

Metric	Value
Accuracy	84.0%

Acknowledgements

• Kaggle Dataset by indk214
• TensorFlow & Keras for deep learning tools

Collaborators

This project is a collaboration between:

• @unknowncurt
• @AmLo
• @Melodiehc

Instructors :

• @scharffc
• @kaleema
• @stephsicilian